The addition of bevacizumab to standard chemotherapy prolonged the median progression free survival (PFS) over chemotherapy alone for women with previously untreated metastatic breast cancer in the pivotal phase III trial, E2100. These findings led to the FDA accelerated approval for bevacizumab in combination with paclitaxel as initial chemotherapy for metastatic disease. The superior PFS in E2100, however, did not translate into an improvement in overall survival (OS) and many patients experienced significant drug-related toxicities. Unfortunately there are no validated biomarkers to help select which patients will experience the optimal benefit to toxicity ratio. We recently demonstrated that vascular endothelial growth factor-A (VEGFA) gene (the target for bevacizumab) amplification and deletion is relatively common in primary breast cancers. We also previously identified two single nucleotide polymorphisms (SNPs) in VEGFA which predicted strongly for an improved median OS and two additional SNPs which predicted protection from significant hypertension when receiving bevacizumab for metastatic breast cancer in E2100. Since that finding, other non-VEGFA SNPs have been correlated with outcome in other randomized phase III trials implementing bevacizumab. Specific Aim #1: Tumor VEGFA amplification or borderline amplification (estimated 14% frequency) will predict superior outcome for patients with metastatic breast cancer receiving bevacizumab in E2100 whereas those with VEGFA deletion (estimated 11% frequency) will predict inferior outcome. We also hypothesize that VEGFA amplification/deletion will not predict outcome in the control arm of E2100. Specific Aim #2: To demonstrate that VEGFA haplotypes and other candidate SNPs will predict superior outcome for patients with metastatic breast cancer receiving bevacizumab in E2100 (but not for the control arm). Specific Aim #3: A combined algorithm calculated from tumor-specific variability (VEGFA amplification/deletion) and host-specific variability (SNPs) will optimally predict outcome (efficacy) with bevacizumab in E2100. Impact: Bevacizumab is a highly active agent in breast cancer but not all patients benefit and there are some substantial toxicities including: stroke and hypertension. Unfortunately there are no validated biomarkers that direct which patients should receive this agent. Additionally, non-selective implementation of bevacizumab has significant negative financial implications on the United States health care system. This proposal has the potential to unveil a predictive signature which will select a subgroup who should receive bevacizumab.